api-discovery
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the Postman CLI to perform authentication checks, search for resources, and query the Postman Context Graph.
- Evidence: Commands such as
postman status,postman search <type> <query>,postman context-graph ask, andpostman dependency addare used throughoutSKILL.mdto interact with the Postman ecosystem. - [DATA_EXFILTRATION]: The skill transmits potentially sensitive task descriptions and queries to an external third-party API service (Orbit) for discovery purposes.
- Evidence:
SKILL.mdandreference/orbit.mdinstruct the agent to send POST requests tohttps://api.buildwithorbit.ai/v1/searchandhttps://api.buildwithorbit.ai/v1/integrate. These requests include task descriptions and optionally identify the agent/model being used via theOrbit-Client-Nameheader. - [EXTERNAL_DOWNLOADS]: The skill fetches external API collections and mocks from Postman's cloud storage to the local development environment.
- Evidence: The
postman dependency addcommand inSKILL.mddownloads selected resources to thepostman/.dependencies/directory and tracks them in.postman/resources.yaml. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and follows integration guidance provided by an external third-party service, which could serve as a vector for malicious instructions if the source were compromised.
- Ingestion points: Responses from
api.buildwithorbit.aiendpoints as described inSKILL.mdandreference/orbit.md. - Boundary markers: Absent; the agent is encouraged to use the returned
STEPSandGOTCHASverbatim to implement API clients. - Capability inventory: The skill has the ability to execute CLI commands and write files to the repository through Postman tooling.
- Sanitization: No explicit validation or sanitization of the remote guidance is implemented before the agent processes it.
Audit Metadata